JOHN WELLS

Goals in teaching English pronunciation

1. Aims in language teaching

The current debate about the phonology of English as an international

language (EIL) should encourage us to think about our aims in
language teaching, and specifically in the teaching of English
pronunciation in the context of English for speakers of other
languages. Some of the questions we need to address are as follows.

• Are we teaching EFL, ESL or EIL? that is, do we intend our

students to use English as a foreign language, as a second
language, or as an international language? Now the mere
formulation of this question exposes its absurdity. English in
Poland may not currently have any role as a second language
in the sense of a role such as it plays in India, Nigeria or
Singapore; but Polish learners of English will surely want to
be able to apply their learning of English both in an EFL
context and in an EIL context. They want to be able to apply
their acquired knowledge of English by participating
wherever English is used. It is not realistic to ask for a choice
between EFL and EIL: our students need both.
• Do you and your students want to be able to interact with
native speakers? or only with non-native speakers? Will
they interact with the British, the Americans, the Australians,
the Irish, the English-speaking West Indians and the
Canadians? Or will they interact with those whose L1 is not
English, for example with the Japanese, the Scandinavians,
 and the Arabs? Or indeed with those who will shortly be your
partners in the European Union – the Italians, the Spanish, the
Austrians – to the extent that they will be speaking English
with them rather than French, German or some other EU
language? Clearly, Polish learners will want to be able to
interact with both native speakers (NSs) and non-native
speakers (NNSs).
• The teaching of English to speakers of other languages may
indeed have different aims in, for example, Britain, Nigeria,
and Japan respectively. In teaching English to immigrants in
Britain, the main aim is clearly to enable learners to interact
with British people, native speakers. In Nigerian primary
schools, it is to enable them to participate in the public life of
their country by interacting with other Nigerians. In Japan a
main focus might indeed be the use of English to
communicate with the Chinese or the Latin Americans.
• What are the student’s personal aims and aspirations in
language learning? Different students in the same class of
school or university may well have rather different aims.
Some just want enough English to communicate at a basic
level, or indeed just enough to pass some examination. Others
aim to achieve the best they possibly can. We must cater for
both types and for those who fall somewhere between.
Speaking personally, I must say that my own aspiration in
learning languages is NS-like proficiency. I acknowledge that
I may be unlikely to attain it. But that doesn’t stop me aiming
for it. I try to inspire my students with the same high ideal. If
it were suggested that I should not even aim so high, I should
feel short-changed.
2. ‘English as an international language’

What, then, are the characteristics of English as an International

Language? Arguably, it suffers from a number of design faults,
characteristics that make it unsuitable for this role newly imposed
upon it.
• It has an elaborate and unwieldy vocabulary. Even among the
most basic and frequent words there are many sets of near-
synonyms such as ill vs. sick, big vs. large, small vs. little,
tricky to distinguish between. Where we have a single noun
king we have three related adjectives: kingly (of Germanic
origin), royal (French) and regal (Latin). They have subtly
different nuances, which may be fine for literature and literary
language, but are a superfluous burden for those who only
want to use the language for practical purposes. The verb
arrive has an associated noun arrival; but for depart the noun
is not *departal but departure. When the plan lands that is not
a *landal or a *landure but a landing; when it takes off again
that is not a *take-offal, *take-offure or *taking-off but a
simple take-off. This inconsistency in derivational
morphology (typical of English) is an unnecessary
complication for NNSs.
• It has a complex syntax, although this is partly compensated
for by the simplicity of the inflectional morphology.
• Its orthography is notoriously inconsistent and irregular. You
cannot safely predict the pronunciation from the spelling. Nor,
given the pronunciation, can you reliably infer the spelling.
• Its phonetics is idiosyncratic, including various
characteristics that are unusual from the point of view of
universals: an large and elaborate vowel system, including
complex processses of length alternation and weakening
(compete-competitive-competition); a consonant system that
includes dental fricatives ([ , ]) and voiced sibilants ([z, ,
 d ]), which are problematic for many learners; words stress
placement that is free, i.e. arbitrary and frequently
unpredictable; and an intonation system that seems to be more
complex and to have a much higher functional load than that
of most other languages.

It it because of such considerations that some (including me) have

argued that for international purposes we ought to use Esperanto,
rather than English. Given, however, that most people seem content
for English to play this role, what special provisions do we need to
make for EIL?

The Lingua Franca Core (LFC) approach can be represented

— with oversimplification and rather unfairly — as saying that we
should ignore the parts of English that NNSs tend to get wrong.
Jennifer Jenkins has made a number of proposals in The Phonology of
English as an International Language (Jenkins 2000). We shall
consider some of them in a moment. If we applied similar proposals
not to phonetics but to grammar, it would arguably mean ignoring
such difficult matters as the articles (coffee—a coffee—the coffee), the
number system (singular vs. plural, dog vs. dogs), the distinction
between countable [C] and uncountable [U] (so that we could happily
talk of informations and furnitures), and the distinction in verbs
between progressive and non-progressive (are you smoking? vs. do
you smoke, which even fluent users of English in Scandinavia
typically ignore). In vocabulary we could stop worrying about false
friends such as actually and eventually, relatively international words
where the NS English meaning is out of line with the meaning in
other languages that have the word.

Many of the oddities of NNS pronunciation of English are due

to inappropriate inference from the spelling. The NS spoken form of
marvellous is [ m (r)v( )l s]. NNSs who say [ mavelus] or the like,
with [u] in the final syllable, are doing so purely on the basis of
(mis)interpreting the spelling. Native speakers pronounce climbing as
[ kla m ] or [ kla m n]. Nigerians who say [ klaimbi g], with [-b-],
do so because of spelling. For NSs, the past tense of look [l k] is
looked [l kt]. Nigerians typically treat the past tense as [d] and then
apply voicing assimilation, giving [lugd]. Arabs speaking English
often treat it as [ d], giving [ l k d]. Arguably, both of these forms are
mispronunciations arising simply from defective teaching: no one has
ever taught such NNSs how the English regular past tense is
pronounced. There is no more reason to regard them as acceptable
than there is for *childs instead of children or *teached instead of
taught.

In cases where NSs make differences in pronunciation that are

not reflected in spelling, NNSs tend to ignore them. Although the
difference between the noun entrance [ entr ns] and the verb to
entrance [ n tr ns] can be coped with, the difference between the verb
to separate [ sep re t] and the adjective separate [ sepr t, sep( )r t]
may be lost, as is that between the verb to document [-ment] and the
noun a document [-m nt]. South and southern have different vowels
for NSs ([sa , s ( )(r)n]), but often not for NNSs. There are many
other ways in which English spelling misleads NNSs, who unlike NSs
learn visually rather than auditorily. NSs pronounce front with the
STRUT vowel (RP [fr nt]); NNSs often use the LOT vowel ([fr nt]),
purely because of the way it is written. There are two possible
remedies for this general problem (if it is indeed a problem): either we
must reform English spelling (and I might mention that I have just
become President of the Simplified Spelling Society) — or teachers of
English to speakers of other languages must teach the pronunciation
of each word as well as its spelling. This implies teaching the use of
phonetic symbols, at least passively for reference.
3. Phonology of EIL?

Jenkins’s proposals still require the mastery of a fair number of

difficult pronunciation points that are not in practice mastered by
many users of EIL.

The consonant [f], a major problem for Koreans, Filipinos and

others. Korean [ph] instead of [f] is likely to trigger a breakdown in
communication, as Jenkins shows; Korean [h ] (their other L1
possibility) is hardly a better substitute. We have to teach the
articulatory difference between bilabial plosive [p] and labiodental
fricative [f]; we have to train the learner not only to produce the
difference but also to perceive it (the latter task being often
overlooked). There is no way to avoid drilling the learner with
minimal pairs such as pork—fork, copy—coffee.

Other consonantal differences that constitute serious problems

for some learners, but which Jenkins rightly insists must be mastered,
include [b–v, r–l, s– , s–z, t –d , j–d ]. Failure to discriminate one or
two of these pairs can perhaps be condoned, given sufficient
redundancy in the context to disambiguate otherwise ambiguous
messages. We can readily cope with Swedish English in which every
/z/ becomes [s], provided that all the rest of the pronunciation is pretty
NS-like. But Japanese English in which [b-v] and [r-l] are confused,
together with various vowel confusions and phonotactic problems,
ends up unintelligible.

I am in favour of Jenkins’s suggestion that l-vocalization

should be allowed, indeed encouraged for those learners for whom
dark /l/ constitutes a problem. There are millions of Londoners and
others who say [m ok] for milk, [b ob] for bulb, [ b to] or [ b o] for
bottle, etc., and I see no reason why the French or the Cantonese
should not do likewise.

Allophonic reduction in vowel length (pre-fortis clipping, as

in right as compared with ride) helps intelligibility, but is difficult to
teach and learn. However phonemic vowel ‘length’ differences,
perhaps better considered primarily as vowel quality differences, are
another matter. Jenkins is right to insist on mastery of the [i – ]
distinction (leave vs. live, sheep vs. ship), which is made by all NSs.
Her wording also implies that the distinctions [u – ] and [ – ] are
equally required, and here I disagree. Millions of Scottish speakers of
English manage perfectly well without any difference between the
vowel of shoot and that of foot, and there are tens of millions of
Americans and Canadians for whom hawk and hock are
homophonous. These distinctions have a low functional load and are
not needed in EIL.

Jenkins’s wording does not leave it entirely clear whether the

vowel oppositions /e-æ, æ- , - / are required in the LFC, but I
assume that they are, despite constituting a considerable problem for
some NNSs. The difficulty with English /æ/ is that many languages
have only two vowels available for the three English vowels /e, æ, /
to be mapped onto. The consequence is that learners disregard either
the /e – æ/ distinction (Polish, Russian, German and Hungarian
learners, who tend to make bed and bad identical) or the /æ - /
distinction (Japanese and Spanish-speaking learners, who tend to
make bad and bud identical). In either case misunderstandings can
result.

It is to be emphasized that we are concerned here with the

vowel system rather than with the details of vowel realization. All
NSs distinguish bed – bad – bud, though the actual vowel qualities
used may vary widely. Listeners can tune in to such variability
without too much difficulty. There are six short vowels in most kinds
of English, representing the standard lexical sets KIT, DRESS, TRAP,
STRUT, LOT, FOOT (Wells 1982:ch. 2), as in bid, bed, bad, bud, cod,
good. There are NS accents that merge STRUT and FOOT (the north of
England, where cut and put rhyme) or TRAP and LOT (popular
Jamaican, where black and block sound identical). But no NS accent
merges DRESS and TRAP (/e - æ/, bed—bad), a distinction that also
bears a high functional load. Nor is there any NS accent that merges
TRAP and STRUT (/æ - /, bad—bud). These oppositions, difficult as
they may be for learners, are ones on which we must insist.

In teaching such vowel oppositions it is important not to

forget to teach the spelling-to-sound rules associated with them. For /e
– æ/ there is a fairly reliable rule: if the spelling is e or ea, the sound
may be /e/ but never /æ/; if the spelling is a, the sound may be /æ/
but never /e/. Thus we have let, dress, when, very, never, dead, bread,
head, pleasure with /e/ and hat, cap, ran, stack, have, gather, tram,
dabble with /æ/. The only exceptions are any and many, together with
ate if pronounced /et/ and the suffix –ary if pronounced /-eri/. For /æ -
/ the rule is 100% reliable: if the spelling is a, the sound may be /æ/
but never /e/; if the spelling is u, o or ou the sound may be / / but
never /æ/. Thus we have hat, cap etc. again with /æ/, and hut, cup,
run, stuck, love, mother, come, touch, trouble with / /.

While there are various NS accents of English that manage

without the opposition between LOT and THOUGHT (don—dawn, RP
/ - /), there are none that dispense with that between THOUGHT and
GOAT (law—low, RP / - /). So here again we must insist that this
distinction be learnt. Again what is important is the systemic contrast
rather than any particular realization: clearly in an EIL context [o ] is
as acceptable for GOAT as [ ] or [o ].

From a comparative and historical perspective, the accents of

England (including RP), Australia, New Zealand, and South Africa are
unusual in having lost the distinction between the lexical sets
THOUGHT and FORCE, merging them as / / in homophones such as
flaw—floor, caught—court, sauce—source (as in Gimson’s joke about
good chefs who, like good journalists, refuse to reveal their / s s z/).
The resultant homophonic clashes do not cause serious problems, even
though Jenkins’s proposals remedy them by restoring historical r as
appropriate.
 Jenkins also insists on the mastery of the [ ] of the NURSE set
(or rather of its rhotic equivalent [ ]). Whichever variant we select,
however, we are dealing with a sound-type that is from the point of
view of language universals highly marked, being vary rare indeed
among the languages of the world – though fortunately, perhaps, the
widely spoken Mandarin Chinese does have a sound similar to
American [ ]. There are many EIL learners for whom this vowel
remains problematic, not least the Japanese, who typically confuse
star and stir.

Let us turn now to the question of phonotactics: cases in

which it is not so much individual sounds that constitute a problem as
their combinations in particular positions in the syllable. Although
Poles have no difficulties with English consonant clusters, there are
many learners who do – Japanese, Chinese, and Koreans for example
as well as speakers of Spanish. Thus an English word such as strong
may come out most easily as [es tron] (Spaniards) or [s t r ]
(Japanese), with the difficult initial consonant cluster /str-/ resolved by
the addition of epenthetic vowels. Rather than add vowels, speakers of
Cantonese tend to omit consonants that are in positions they find
difficult, which has an even worse effect on intelligibility. In word-
initial position the clusters in such everyday words such as pray,
bread, train, queen, splash may offer a problem; so in word-final
position may the clusters in milk, lamp, left, fox and wasp (not to
mention its plural wasps).

Voiced obstruents are not a problem for speakers of Polish,

German, or Russian, but producing them in word-final position is.
Hence they must learn to produce voiced (or at least lenis) obstruents
in such words as rub, bad, big, love, rose, rage. Whether bed is
pronounced as [bet] (German) or [be ] (Cantonese), in each case the
NS opposition between final /d/ and /t/ is lost.

A particular problem with consonants is that the L1 may have

phonological processes – allophonic or assimilatory – that are
inappropriate in English. However learners of English will tend to
apply them in English unless taught not to. Thus Korean learners, for
example, need to be warned against the Korean assimilatory processes
that turn pop music into po[m] music or Rugrats into Ru[ n]ats. Poles
should be discouraged from applying Polish-style voicing assimilation
such as makes ice dancing sound like eyes dancing and pick them up
like pig them up.

When we turn to suprasegmental matters, Jenkins rightly

insists on the importance of not accenting function words. There must
be a difference between a big one (e.g. when we are talking about
waves, a big wave: one is a function word) and a big one (which
might be a large figure one: one is a form word). She rightly insists
also on the importance of deaccenting repeated lexical items, or of
lexical items with the same semantic referent. Although this principle
applies in many other languages more or less as in English, there are
differences of detail: as pointed out by Ortiz-Lira, 1995, where the
English reply to We’re already late might be I don’t care if we are
late, with the repeated word late deaccented, the Spanish equivalent
would be Pero si ya estamos atrasados – No me importa si estamos
atrasados with no such change in accent pattern.

In summary, my prioritizing recommendations for the teaching of

English pronunciation in an EFL/EIL context would be:
• to concentrate on the matters that most impede intelligibility;
while encouraging fluency and confidence;
• not to neglect the need to interact with NSs; arguably, we also
need to educate the NSs;
• to exploit the findings of contrastive analysis to help pinpoint
likely areas of difficulty.

While contrastive analysis does not provide all the answers, it

goes a good way towards pointing us in the right direction. This
means, for instance, that Polish learners of English must pay particular
attention to those consonants that are not found (or not found as
phonemes, or found with very different phonetic realization) in their
L1: / , , , r, h/; to final obstruent vocing, and to aspiration; among
English vowels, to /æ, , , , e /, to pre-fortis clipping, to vowel
duration and to weakening.
Thank you [ æ k ju], or as we might say in the LFC [ te k ju veri
mat ].

References

Jenkins, Jennifer, 2000. The Phonology of English as an International

Language. Oxford University Press.
Lewis, Jack Windsor (ed.), 1995. Studies in General and English
Phonetics. London: Routledge.
Ortiz-Lira, Héctor, 1995. ‘Nucleus placement in English and Spanish:
a pilot study’. In Lewis 1995. Wells, J.C., 1982. Accents of
English. Three volumes. Cambridge University Press.